Simple Algorithm
Quantum Computing · Rydberg State · 30 min

States as Rays: Projective Hilbert Space

Global phase, CP^n and the Fubini–Study metric.

∣ψ⟩\ket\psi and −∣ψ⟩-\ket\psi are different vectors but the same physical state. Take that seriously and the state space stops being a vector space. It becomes complex projective space, a curved manifold with its own natural distance, its own symmetries and its own speed limit.

What you’ll learn

  • Define pure states as rays and identify the state space with CPd−1\mathbb{CP}^{d-1}, with CP1≅S2\mathbb{CP}^1\cong S^2.
  • Compute Fubini–Study distances and relate them to fidelity, trace distance and Bloch angles.
  • State Wigner's theorem and explain why antiunitary maps such as the universal NOT are unphysical.
  • Derive the Anandan–Aharonov speed ds/dt=ΔE/ℏds/dt=\Delta E/\hbar and the Mandelstam–Tamm bound.

From vectors to rays

Every prediction depends only on rays: Pr⁡(ϕ∣ψ)=∣⟨ϕ∣ψ⟩∣2/(⟨ϕ∣ϕ⟩⟨ψ∣ψ⟩)\Pr(\phi\mid\psi)=\abs{\braket{\phi|\psi}}^2/\big(\braket{\phi|\phi}\braket{\psi|\psi}\big) and ⟨A⟩=⟨ψ|A|ψ⟩/⟨ψ∣ψ⟩\expval{A}=\mel{\psi}{A}{\psi}/\braket{\psi|\psi}. For d=2d=2, CP1\mathbb{CP}^1 is the Bloch sphere S2S^2: unit vectors form S3⊂C2S^3\subset\C^2, and forgetting the phase is the Hopf fibration S3→S2S^3\to S^2, whose fibers are circles of global phase. For nn qubits, CP2n−1\mathbb{CP}^{2^n-1} has real dimension 2n+1−22^{n+1}-2.

Global phase is unobservable, but a phase is only global relative to a system. Apply eiγe^{i\gamma} conditionally on a control qubit in superposition and it becomes a relative phase of the larger system, hence measurable. That is phase kickback, the engine of phase estimation.

The Fubini–Study metric

DerivationThe Bloch sphere has radius one half
  1. ⟨dψ∣dψ⟩=14dθ2+sin⁡2θ2 dφ2,⟨ψ∣dψ⟩=isin⁡2θ2 dφ\braket{d\psi|d\psi}=\tfrac14d\theta^2+\sin^2\tfrac\theta2\,d\varphi^2,\qquad \braket{\psi|d\psi}=i\sin^2\tfrac\theta2\,d\varphi

    Parametrize ∣ψ⟩=cos⁡θ2∣0⟩+eiφsin⁡θ2∣1⟩\ket\psi=\cos\tfrac\theta2\ket0+e^{i\varphi}\sin\tfrac\theta2\ket1 and differentiate:

  2. ds2=14dθ2+sin⁡2θ2(1−sin⁡2θ2)dφ2=14(dθ2+sin⁡2θ dφ2)ds^2=\tfrac14d\theta^2+\sin^2\tfrac\theta2\Big(1-\sin^2\tfrac\theta2\Big)d\varphi^2=\tfrac14\big(d\theta^2+\sin^2\theta\,d\varphi^2\big)

    Subtract the phase direction:

  3. dFS=Θ2,Area⁡(CP1)=4π(12)2=πd_{FS}=\tfrac\Theta2,\qquad \operatorname{Area}\big(\mathbb{CP}^1\big)=4\pi\big(\tfrac12\big)^2=\pi

    That is the round metric of a sphere of radius 12\tfrac12: FS distances are half the Bloch angles Θ\Theta, i.e. ∣⟨ψ∣ϕ⟩∣=cos⁡Θ2\abs{\braket{\psi|\phi}}=\cos\tfrac\Theta2. Orthogonal states are antipodal (Θ=π\Theta=\pi) at FS distance π2\tfrac\pi2.

The FS distance packages the other pure-state measures: fidelity F=cos⁡2dFSF=\cos^2d_{FS} and trace distance D=sin⁡dFSD=\sin d_{FS}. Up to scale it is the unique Riemannian metric on CPd−1\mathbb{CP}^{d-1} invariant under all unitaries, and in this normalization Vol⁡(CPn)=πn/n!\operatorname{Vol}(\mathbb{CP}^n)=\pi^n/n!. The normalized volume is the Haar measure on pure states used in every random-state argument.

Predict before you look

What is the Fubini–Study distance between ∣0⟩\ket0 and ∣+⟩\ket+, in degrees? (Their Bloch vectors are 90∘90^\circ apart.)

Show the answer

45 °

arccos⁡∣⟨0∣+⟩∣=arccos⁡12=45∘\arccos\abs{\braket{0|+}}=\arccos\tfrac1{\sqrt2}=45^\circ, half the Bloch angle.

Symmetries: Wigner's theorem

DerivationProof sketch
  1. T[ek]=[ek′],⟨ej′∣ek′⟩=δjkT[e_k]=[e_k'],\qquad \braket{e_j'|e_k'}=\delta_{jk}

    Fix an orthonormal basis {∣ek⟩}\{\ket{e_k}\}. Transition probabilities 00 and 11 are preserved, so representatives ∣ek′⟩\ket{e_k'} of the rays T[ek]T[e_k] again form an orthonormal basis.

  2. T[e1+ek]=[e1′+ek′]T\big[e_1+e_k\big]=\big[e_1'+e_k'\big]

    The image of [e1+ek][e_1+e_k] has overlap 12\tfrac12 with [e1′][e_1'] and [ek′][e_k'] and 00 with the rest, so it is [e1′+eiαek′][e_1'+e^{i\alpha}e_k']. Absorb eiαe^{i\alpha} into ∣ek′⟩\ket{e_k'}:

  3. ∣ck′∣=∣ck∣,∣1+ck′∣=∣1+ck∣⟹ck′∈{ck, ck∗}\abs{c_k'}=\abs{c_k},\quad\abs{1+c_k'}=\abs{1+c_k}\quad\Longrightarrow\quad c_k'\in\{c_k,\,c_k^*\}

    For ψ=∑kckek\psi=\sum_kc_ke_k with c1=1c_1=1, the image is [∑kck′ek′]\big[\sum_kc_k'e_k'\big] with c1′=1c_1'=1 and ∣ck′∣=∣ck∣\abs{c_k'}=\abs{c_k}; preserving the overlap with [e1+ek][e_1+e_k] forces ∣1+ck′∣=∣1+ck∣\abs{1+c_k'}=\abs{1+c_k}:

  4. U(∑kckek)=∑kckek′or∑kck∗ek′U\Big(\sum_kc_ke_k\Big)=\sum_kc_ke_k'\qquad\text{or}\qquad\sum_kc_k^*e_k'

    Testing against rays such as [e1+ej+ek][e_1+e_j+e_k] and [e1+ej+iek][e_1+e_j+ie_k] shows that the choice cannot vary with kk or with ψ\psi. So TT is induced either by a unitary or by a unitary composed with complex conjugation KK:

Antiunitaries do occur: time reversal is antiunitary. But continuous dynamics must be unitary: a product of two antiunitaries is unitary, and each element of a continuous one-parameter group is a square, V(t)=V(t/2)2V(t)=V(t/2)^2. On the Bloch sphere, unitaries act as rotations (PU(2)≅SO(3)PU(2)\cong SO(3)), while antiunitaries act as orientation-reversing isometries such as reflections; complex conjugation KK maps (x,y,z)↦(x,−y,z)(x,y,z)\mapsto(x,-y,z).

Check your understanding

The "universal NOT" sends every qubit pure state to its orthogonal state, i.e. reflects the Bloch sphere through its center. What does Wigner's theorem say about it?

  1. It is the unitary YY
  2. It is antiunitary (∣ψ⟩↦YK∣ψ⟩\ket\psi\mapsto YK\ket\psi), so no physical device implements it exactly
  3. It does not preserve transition probabilities
  4. It is the unitary XZXZ, up to a global phase
Show the answer

B. It is antiunitary (∣ψ⟩↦YK∣ψ⟩\ket\psi\mapsto YK\ket\psi), so no physical device implements it exactly

The point reflection r⃗↦−r⃗\vec r\mapsto-\vec r preserves ∣⟨ψ∣ϕ⟩∣2=12(1+r⃗⋅s⃗)\abs{\braket{\psi|\phi}}^2=\tfrac12(1+\vec r\cdot\vec s) but has determinant −1-1, so it is not a rotation; Wigner forces it to be antiunitary. Explicitly YKYK maps α∣0⟩+β∣1⟩\alpha\ket0+\beta\ket1 to −iβ∗∣0⟩+iα∗∣1⟩-i\beta^*\ket0+i\alpha^*\ket1, which is orthogonal to the input. Only approximate versions exist, with optimal fidelity 23\tfrac23.

Geometry in motion: speed limits and geometric phase

Curves in CPd−1\mathbb{CP}^{d-1} have lengths, so dynamics has a speed. Under iℏddt∣ψ⟩=H∣ψ⟩i\hbar\tfrac{d}{dt}\ket\psi=H\ket\psi, the speed in the FS metric is set by the energy uncertainty ΔE=⟨H2⟩−⟨H⟩2\Delta E=\sqrt{\expval{H^2}-\expval{H}^2}:

DerivationAnandan–Aharonov speed and the Mandelstam–Tamm bound
  1. ds2=dt2ℏ2(⟨H2⟩−⟨H⟩2)=ΔE2ℏ2 dt2ds^2=\frac{dt^2}{\hbar^2}\Big(\expval{H^2}-\expval{H}^2\Big)=\frac{\Delta E^2}{\hbar^2}\,dt^2

    Insert ∣dψ⟩=−iℏH∣ψ⟩ dt\ket{d\psi}=-\tfrac i\hbar H\ket\psi\,dt into ds2=⟨dψ∣dψ⟩−∣⟨ψ∣dψ⟩∣2ds^2=\braket{d\psi|d\psi}-\abs{\braket{\psi|d\psi}}^2:

  2. dFS(ψ(0),ψ(t))≤∫0tΔE(t′)ℏ dt′d_{FS}\big(\psi(0),\psi(t)\big)\le\int_0^t\frac{\Delta E(t')}{\hbar}\,dt'

    So the state moves through projective space at speed ΔE/ℏ\Delta E/\hbar (Anandan & Aharonov, 1990). The distance travelled is at most the path length:

  3. t⊥ ≥ πℏ2 ΔE(Mandelstam–Tamm, 1945)t_\perp\ \ge\ \frac{\pi\hbar}{2\,\Delta E}\qquad\text{(Mandelstam–Tamm, 1945)}

    For time-independent HH, ΔE\Delta E is conserved, and reaching an orthogonal state means covering FS distance π2\tfrac\pi2:

  4. t⊥ ≥ max⁡{πℏ2ΔE, πℏ2(⟨H⟩−E0)}t_\perp\ \ge\ \max\Big\{\frac{\pi\hbar}{2\Delta E},\ \frac{\pi\hbar}{2\big(\expval{H}-E_0\big)}\Big\}

    Margolus and Levitin (1998) proved a complementary bound in terms of the mean energy above the ground energy E0E_0. Both are tight for an equal superposition of two energy levels:

Closed loops carry a second geometric quantity. Transport a state around a loop in CPd−1\mathbb{CP}^{d-1} while keeping ⟨ψ∣dψ⟩=0\braket{\psi|d\psi}=0 (no local phase change) and it returns multiplied by the holonomy of the Hopf bundle, the Aharonov–Anandan geometric phase. For a qubit its magnitude is half the solid angle enclosed on the Bloch sphere, independent of how fast the loop is traversed (see Geometric (Berry) Phase).

Key takeaways

  • Pure states are rays: the state space of Cd\C^d is CPd−1\mathbb{CP}^{d-1}, of real dimension 2d−22d-2; for a qubit the Hopf fibration S3→S2S^3\to S^2 quotients out the global phase.
  • Global phase is unobservable, but a conditional phase is relative and measurable (phase kickback).
  • The Fubini–Study distance arccos⁡∣⟨ψ∣ϕ⟩∣\arccos\abs{\braket{\psi|\phi}} is half the Bloch angle, with F=cos⁡2dFSF=\cos^2d_{FS} and D=sin⁡dFSD=\sin d_{FS}.
  • Wigner: symmetries of rays are unitary or antiunitary; continuous dynamics is unitary, and the universal NOT is not physical.
  • States move at FS speed ΔE/ℏ\Delta E/\hbar, which gives the Mandelstam–Tamm bound t⊥≥πℏ/(2ΔE)t_\perp\ge\pi\hbar/(2\Delta E).

Check yourself

In the interactive lesson these are scored with your confidence, and what you miss comes back for review.

Question 1

What is the real dimension of the pure-state space of nn qubits?

  1. 2n+12^{n+1}
  2. 2n−12^n-1
  3. 2n+1−22^{n+1}-2
  4. 2n2n
Show the answer

C. 2n+1−22^{n+1}-2

C2n\C^{2^n} has 2n+12^{n+1} real dimensions; normalization removes one and global phase another, so dim⁡RCP2n−1=2(2n−1)\dim_\R\mathbb{CP}^{2^n-1}=2(2^n-1). For n=1n=1 that is 22, the Bloch sphere.

Question 2

What is the Fubini–Study distance between two orthogonal pure states?

  1. π\pi
  2. 2\sqrt2
  3. 11
  4. π2\tfrac\pi2
Show the answer

D. π2\tfrac\pi2

arccos⁡∣⟨ψ∣ϕ⟩∣=arccos⁡0=π2\arccos\abs{\braket{\psi|\phi}}=\arccos0=\tfrac\pi2, the maximum. In this metric the Bloch sphere has radius 12\tfrac12, and orthogonal states are antipodal: half a great circle of circumference π\pi.

Question 3

A qubit evolves under a time-independent Hamiltonian, and its initial state has energy uncertainty ΔE=ℏΩ/2\Delta E=\hbar\Omega/2. What is the shortest possible time to reach an orthogonal state?

  1. π/Ω\pi/\Omega
  2. π/(2Ω)\pi/(2\Omega)
  3. 2π/Ω2\pi/\Omega
  4. 1/Ω1/\Omega
Show the answer

A. π/Ω\pi/\Omega

Mandelstam–Tamm: t⊥≥πℏ/(2ΔE)=π/Ωt_\perp\ge\pi\hbar/(2\Delta E)=\pi/\Omega, attained by the resonant π\pi-pulse.

Question 4

For unit vectors ∣ψ⟩\ket\psi, which of these is not a function of the ray [ψ][\psi] alone?

  1. ∣⟨0∣ψ⟩∣2\abs{\braket{0|\psi}}^2
  2. ⟨ψ|Z|ψ⟩\mel{\psi}{Z}{\psi}
  3. The complex number ⟨0∣ψ⟩\braket{0|\psi}
  4. The projector ∣ψ⟩ ⁣⟨ψ∣\dyad\psi
Show the answer

C. The complex number ⟨0∣ψ⟩\braket{0|\psi}

Replacing ∣ψ⟩\ket\psi by eiγ∣ψ⟩e^{i\gamma}\ket\psi multiplies ⟨0∣ψ⟩\braket{0|\psi} by eiγe^{i\gamma}. Probabilities, expectation values and ∣ψ⟩ ⁣⟨ψ∣\dyad\psi are phase-invariant. Ratios such as ⟨1∣ψ⟩/⟨0∣ψ⟩\braket{1|\psi}/\braket{0|\psi} are ray invariants: they are the projective coordinates on CP1\mathbb{CP}^1 that encode the relative phase.

Further reading

  • I. Bengtsson & K. Życzkowski, Geometry of Quantum States: An Introduction to Quantum Entanglement (Cambridge, 2nd ed., 2017).
  • V. Bargmann, "Note on Wigner's theorem on symmetry operations," J. Math. Phys. 5, 862 (1964).
  • J. Anandan & Y. Aharonov, "Geometry of quantum evolution," Phys. Rev. Lett. 65, 1697 (1990).
  • L. Mandelstam & I. Tamm, "The uncertainty relation between energy and time in non-relativistic quantum mechanics," J. Phys. USSR 9 (1945).
  • N. Margolus & L. B. Levitin, "The maximum speed of dynamical evolution," Physica D 120, 188 (1998).